TESTING A CONCEPTUAL MODEL OF VOCAL TREMOR: RESPIRATORY
AND LARYNGEAL CONTRIBUTIONS TO ACOUSTIC MODULATION
by
Jordon LeBaron
A thesis submitted to the faculty of The University of Utah
in partial fulfillment of the requirements for the degree of
Master of Science
in
Speech-Language Pathology
Department of Communication Sciences and Disorders
The University of Utah
December 2016
Copyright © Jordon LeBaron 2016
All Rights Reserved
T h e U n i v e r s i t y o f U t a h G r a d u a t e S c h o o l
STATEMENT OF THESIS APPROVAL
The thesis of Jordon LeBaron
has been approved by the following supervisory committee members:
Julie Barkmeier-Kraemer , Chair May 4, 2016
Date Approved
Michael Blomgren , Member May 4, 2016
Date Approved
Bruce Smith , Member May 4, 2016
Date Approved
and by Michael Blomgren , Chair/Dean
Department/College/School of Communication Sciences and Disorders
and by David B. Kieda, Dean of The Graduate School.
ABSTRACT Vocal tremor is a neurogenic voice disorder characterized by rhythmic
modulation of pitch and loudness during sustained phonation and is acoustically
measured as modulation of formants, fundamental frequency (fo), and sound
pressure level (SPL). To date, links between oscillating vocal tract structures and
acoustic modulation of the first two formants were shown in those with vocal
tremor. However, laryngeal and respiratory contributions to acoustic modulation
patterns in those with vocal tremor are difficult to separate. The purpose of this
study was to compare acoustic patterns associated with volitional laryngeal
versus respiratory structure oscillations in trained singers. Laryngeal oscillation
was hypothesized to correspond with fo modulation patterns, whereas respiratory
system oscillation was hypothesized to correspond with SPL modulation. Ten
classically trained female singers with no less than 5 years’ experience and no
history or current complaints of voicing problems were recruited between 40–65
years of age. All participants underwent simultaneous recording of
nasoendoscopic views of the larynx, respiratory kinematic and acoustic signals
during three trials of sustained phonation of /i/ using either vibrato or the
Accented Method of Voicing (AMV). Normalized measures of signal modulation
rate and magnitude were completed on the acoustic (fo and SPL) and kinematic
recordings. A mixed effects logistic regression compared within subject
measurement differences between voicing conditions. The results showed
iv
significantly greater magnitude of respiratory kinematics during AMV (47.5%
(+1.2)) than for vibrato (0% (+0) (p < .001) corresponding with significantly
greater SPL modulation magnitude (AMV = 40% (+20); vibrato = (10% (+0)),
respectively (p = .026). A significant difference was also found between voicing
conditions for modulation rate of fo (p = .049) and SPL (p < .001). The rates of
modulation during AMV were slower (fo = 2.8 (+ .8) Hz; SPL = 2.1 (+ .7) Hz) than
for vibrato (fo = 5.1 (+ .7) Hz; SPL = 5 (+ .6) Hz). However, laryngeal kinematic
and acoustic fo and SPL magnitude patterns did not differ between voicing
conditions. Outcomes support predicted contributions of the respiratory system
to voicing modulation; however, the larynx appears interactive with the
respiratory and other speech structures during voicing.
TABLE OF CONTENTS
ABSTRACT ...........................................................................................................iii LIST OF FIGURES ...............................................................................................vii LIST OF TABLES ..................................................................................................ix ACKNOWLEDGEMENTS ......................................................................................x INTRODUCTION ...................................................................................................1
Background .....................................................................................................1 Statement of Purpose .....................................................................................14
METHODS ...........................................................................................................16
Subject ............................................................................................................16 Screening .......................................................................................................16 Procedures .....................................................................................................17 Respiratory Kinematic Recordings ...............................................................17
Respiratory Kinematic Equipment .............................................................17 Respiratory Kinematic Signal Calibration ..................................................19 Respiratory Kinematic Procedures ............................................................20
Acoustic Recordings .......................................................................................20 Acoustic Recording Equipment .................................................................20 Acoustic Recording Procedures ................................................................20
Laryngeal Imaging ..........................................................................................22 Laryngeal Imaging Equipment ..................................................................22 Laryngeal Imaging Procedures .................................................................22 DATA COLLECTION AND MEASURES ..............................................................24
Classification of Participant Voice Modulation Conditions by Expert Judges ............................................................................................................24 Hearing Screenings ........................................................................................24 Listening Task ................................................................................................24 Physiologic Measurement of all Recordings ...................................................26 Respiratory Kinematic Analysis ......................................................................28
Respiratory Kinematic Oscillation Rate .....................................................28
vi
Respiratory Kinematic Measure Adjustments for Slope ............................28 Respiratory Kinematic Oscillation Extent ..................................................32
Acoustic Measures .........................................................................................33 Acoustic Modulation Rate .........................................................................33 Acoustic Modulation Extent .......................................................................34
fo Extent Measures ....................................................................................35 SPL Extent Measures ...............................................................................35
Laryngeal Imaging Kinematic Analysis ...........................................................36 Laryngeal Oscillation Rate .......................................................................37 Laryngeal Oscillation Extent .....................................................................37 Referent Anatomical Distance Measures .......................................37
Laryngeal Lengthwise Extent Measures ........................................38 Laryngeal Abduction/Adduction Extent Measures ..........................38
Statistical Analysis ..........................................................................................39 Intrarater Reliability ...................................................................................41 RESULTS ............................................................................................................42
Qualitative Analysis of Results .......................................................................42 Respiratory System Contributions to Acoustic Modulation .............................42 Phonatory System Contributions to Acoustic Modulation ...............................44 Vocal Tract Movements by Condition .............................................................47
Vertical Laryngeal Movement ....................................................................47 Pharyngeal Movement ..............................................................................48
DISCUSSION .......................................................................................................49
Respiratory System Contributions to Acoustic Modulation .............................50 Laryngeal Contributions to Acoustic Modulation ............................................52 Application of Current Findings to Vocal Tremor ............................................55 Limitations ......................................................................................................56 Conclusion ......................................................................................................58
APPENDIX ...........................................................................................................59 REFERENCES ....................................................................................................61
LIST OF FIGURES
Figures
1 Barkmeier-Kraemer and Story (2010) Conceptual Model of Vocal Tremor .......7
2 Lengthwise (A) and medial/lateral (B) oscillation directions within the Phonatory System ..............................................................................................8
3 LabChart Pro display of simultaneously recorded laryngeal images and acoustic and respiratory signals .......................................................................18
4 A comparison of signals rated for each voicing condition (i.e., vibrato and AMVT). A. A comparison of signals rated for the vibrato condition from the two subjects rated with 44% expert agreement from S03 (a-c) and S04 (d-f) and signals with 100% expert agreement from S05 (g-i) versus. B A comparison of signals rated for the AMVT condition with 67% expert agreement from S10 (a-c) and 100% expert agreement from S05 (d-f) ............................................27 5 Example of determining rate for respiratory kinematic oscillation. Each arrow marks the peak of each modulation cycle displayed in the 2-s window. A total of 11 peak-to-peak cycles are shown in the 2-s window giving a 5.5 Hz respiratory kinematic rate ........................................................29
6 Example of measuring extent of respiratory kinematic oscillation. The minimum and maximum values associated with summated rib cage and abdominal movements in the figure represent the original data points. The original data points were corrected before extent was calculated. After adjustment for the sloping values, the relative %VC extent was calculated as described above ...............................................................................................29
7 Example of fo (top line) and SPL (bottom line) plot of vibrato within Praat. Arrows have been added to the signal to indicate the beginning of the
cycle .................................................................................................................32
8 Example of fo (top line) and SPL (bottom line) plot of vibrato within Praat. Arrows have been added to the signal to indicate the maximum and minimum points of cycle 3 (fo) and cycle 8 (SPL) ............................................................32
viii
9 An example of laryngeal imaging measures for the first end point of one laryngeal oscillatory cycle. Each panel displays A) the image analyzed, B) the anatomical referent line measure, C) the relative measure of vocal fold length, and D) the relative measure of interarytenoid distance associated with the first end point of one laryngeal oscillation cycle. B. Example of laryngeal imaging measures for the second end point of one laryngeal oscillatory cycle. Each panel displays A) the image analyzed, B) the anatomical referent line measure, C) the relative measure of vocal fold length, and D) the relative measure of interarytenoid distance associated with the second end point of one laryngeal oscillation cycle ........................................................................36
10 Kinematic extent comparisons between voicing conditions (i.e., AMVT and vibrato) ..........................................................................................................45 11 Kinematic extent comparisons between voicing types (i.e., AMVT and vibrato) ..........................................................................................................45
12 Acoustic extent comparisons between voicing conditions (i.e., AMVT and vibrato) ..........................................................................................................46
13 Acoustic rate comparisons between voicing conditions (i.e., AMVT and vibrato) .........................................................................................................46
LIST OF TABLES
Tables 1 Speech structures reported to exhibit tremor ....................................................4
2 Barkmeier-Kraemer and Story (2010) summary of conceptual model of vocal tremor ................................................................................................................9 3 Aims of this study ...........................................................................................15
4 Percent agreement between expert judges and intended production condition .........................................................................................................26
5 Intrarater reliability determined using intraclass correlations .........................40
6 AVMT voicing condition descriptive statistical summary for individual subjects ...........................................................................................................43
7 Vibrato voicing condition descriptive statistical summary for individual subjects ...........................................................................................................43
ACKNOWLEDGEMENTS
This investigation was supported by the University of Utah Study Design
and Biostatistics Center, with funding in part from the National Center for
Research Resources and the National Center for Advancing Translational
Sciences, National Institutes of Health, through Grant 8UL1TR000105 (formerly
UL1RR025764).
The completion of thesis would not be possible without the guidance, help,
and support from Dr. Julie Barkmeier-Kraemer, Dr. Michael Blomgren, Dr. Bruce
Smith, Dr. Ingo Titze, Dr. Angela Presson, Chong Zhang, Dan Houtz, Elena
Cervantes, Amy Collard, and Sharon Benavides. A sincere expression of
gratitude is owed to my thesis committee supervisor, Dr. Barkmeier-Kraemer, for
taking an active role in my research education and providing constant guidance
and support.
Finally, this thesis and my entire graduate education would not be possible
without the loving support of my wife, Angela, who made it possible for me to
work long hours, and to whom I owe an enormous amount.
INTRODUCTION
Background
Tremor is defined as involuntary, rhythmic, oscillatory movement produced
by either synchronous or alternating contractions of antagonistic muscles (e.g.,
the biceps and triceps) (Dalvi & Premkumar, 2011; Finnegan, Luschei,
Barkmeier, & Hoffman, 2003; Schneider & Deuschl, 2015). Tremor is considered
to occur related to disinhibition, excitation, or poor regulation of central nervous
system oscillatory neural networks associated with neurological disorders,
pharmaceutical side effects, or limbic system effects (Dalvi & Premkumar, 2011).
Tremor is typically classified by its rate, the structures affected, and conditions
under which it manifests (Dalvi & Premkumar, 2011).
The majority of the literature describing tremor focuses on neurological
disorders involving the limbs, head, and trunk of the body. However, tremor can
also involve structures associated with speaking, typically resulting in the
production of a shaky voice quality referred to as, “vocal tremor.” Chronic
occurrence or progression of vocal tremor has been shown to significantly impact
the intelligibility and quality of life of affected individuals (Louis & Machado,
2015). The resulting negative impact on quality of life may motivate them to seek
remediation through pharmaceutical (Gurey, Sinclair, Blitzer, 2013; Warrick et al.,
2000), behavioral (Barkmeier-Kraemer, Lato, & Wiley, 2011) or surgical
management (Taha, Janszen, & Favre, 1999). Current knowledge about vocal
2
tremor is related to its association with other neurological disorders and its
acoustic patterns. The most common etiology associated with vocal tremor is
Essential Tremor (ET), the most common form of movement disorder (Dalvi &
Premkumar, 2011). Interestingly, 93% of those diagnosed with Essential Vocal
Tremor are female (Sulica & Louis, 2010). Vocal tremor has also been identified
in individuals diagnosed with other neurological disorders such as spasmodic
dysphonia and Parkinson’s disease (Wolraich, Marchis-Cristan, Redding, Khella,
& Mirza, 2010); however, demographic comparisons of vocal tremor across the
latter neurological disorders have not been completed.
A small number of studies have characterized speaking patterns
associated with vocal tremor (Lundy, Roy, Xue, Casiano, & Jassir, 2004) as well
as the ability to perceive vocal tremor across different speech contexts (Brown &
Simonson, 1963; Lederle, Barkmeier-Kraemer, & Finnegan, 2012). Vocal tremor
is best perceived during sustained phonation context (Brown & Simonson, 1963),
although severe vocal tremor can also be perceived within connected speech
(Lederle et al., 2012). Individuals with vocal tremor have also been shown to
speak at a slower rate, on average, compared to normal speakers (Lundy et al.,
2004). In general, vocal tremor is perceived as a rhythmic modulation of pitch
and loudness associated with acoustic modulation of fundamental frequency (fo)
and sound pressure level (SPL), respectively.
Modulation of pitch and loudness associated with vocal tremor has been
characterized by measuring the acoustic correlates, fundamental frequency (fo)
and sound pressure level (SPL), respectively. Vocal tremor modulation rate for
both fo and SPL has been reported to occur between 3-8 Hz (Brown & Simonson,
3
1963; Dromey, Warrick, & Irish, 2002; Ramig & Shipp, 1987). The extent of fo
modulation in vocal tremor has been reported to range between 3-17% with an
extent of SPL modulation reported to range between 19-61% (Barkmeier-
Kraemer, Lato, & Wiley, 2011; Dromey et al., 2002; Ramig & Shipp, 1987).
Thus, it appears that the larger extent of acoustic modulation during vocal tremor
may be due to SPL compared to fo modulation.
The majority of literature addressing vocal tremor has characterized the
rate, extent, and conditions under which it is detected relying primarily on
associated acoustic patterns. A small number of studies also investigated
musculature or structures within the speech mechanism exhibiting tremor
associated with vocal tremor. Based on prior literature identifying tremor within
structures of the speech mechanism, the majority noted tremor within the
pharyngeal constrictors (Sulica & Louis, 2010), larynx (Ackermann & Ziegler,
1991; Adler et al., 2004; Bové et al., 2006; Finnegan et al., 2003; Gamboa et al.,
1998; Sulica & Louis, 2010; Tomoda, Shibasaki, Kuroda, & Shin, 1987), and
tongue (Gamboa et al., 1998; Jiang, Lin, & Hanson, 2000; Sulica & Louis, 2010;
Lester, Barkmeier-Kraemer, & Story, 2013) (see Table 1). The latter structures
may be most frequently associated with vocal tremor due to their ease of
observation during endoscopic evaluation. However, tremor has also been
identified in other structures such as the soft palate (Sulica & Louis, 2010) and
respiratory musculature (Tomoda et al., 1987). Although the majority of literature
addressing vocal tremor has focused on the larynx, approximately 25% of those
with vocal tremor exhibit tremor in structures outside of the larynx, or within the
vocal tract (Bové et al., 2006). Although prior studies have noted tremor affecting
4
Table 1. Speech structures reported to exhibit tremor.
Study Subjects
Methods Speech Structures Studied with Tremor
Sulica, L., & Louis, E. D. (2010). Clinical characteristics of essential voice tremor: A study of 34 cases. The Laryngoscope, 120(3), 516-528.
N=34 with ET
Vocal Tremor Scoring System (VTSS) for laryngeal and pharyngeal movement, Washington Heights Inwood Genetic Study of Essential Tremor (WHIGET) rating scale for arm tremor, Voice Handicap Index (VHI) for voice disability rating.
Larynx, Pharynx, Palate, Tongue
Finnegan, E. M., Luschei, E. S., Barkmeier, J. M., & Hoffman, H. T. (2003). Synchrony of laryngeal muscle activity in persons with vocal tremor. Archives of Otolaryngology–Head & Neck Surgery, 129(3), 313-318.
N=6 with VT
EMG and acoustic analysis from voice recordings
Laryngeal musculature: Cricothyroid, Thyroarytenoid, Sternohyoid, Thyrohyoid
Tomoda, H., Shibasaki, H., Kuroda, Y., & Shin, T. (1987). Voice tremor: Dysregulation of voluntary expiratory muscles. Neurology, 37(1), 117-122.
N=3 with Tremor in voice and hands
EMG and acoustic analysis from voice recordings
Cricothyroid (larynx), Rectus abdominis (chest wall)
Ackermann, H., & Ziegler, W. (1991). Cerebellar voice tremor: an acoustic analysis. Journal of Neurology, Neurosurgery & Psychiatry, 54(1), 74- 76.
N=1 with Cerebellar tremor
Acoustic analysis from voice recordings
Larynx only
Gamboa, J., Jiménez-Jiménez, F. J., Nieto, A., Cobeta, I., Vegas, A., Ortí-Pareja, M., García-Albea, E. (1998). Acoustic voice analysis in patients with essential tremor. Journal of Voice, 12(4), 444-452.
N=56 (28 with ET; 28 control)
Acoustic analysis from voice recordings
Larynx, Vocal Tract (articulators)
Jiang, J., Lin, E., & Hanson, D. G. (2000). Acoustic and Airflow Spectral Analysis of Voice Tremor. Journal of Speech, Language & Hearing Research, 43(1), 191.
N=10 (5M, 5F) neuro-logical disease showing signs of VT
Acoustic analysis, airflow analysis Vocal Tract (articulators)
5
Table 1. Continued
Lester, R. A., Barkmeier-Kraemer, J., & Story, B. H. (2013). Physiologic and Acoustic Patterns of Essential Vocal Tremor. Journal of Voice, 27(4), 422-432.
N=1 with EVT Rigid videostroboscopy, acoustic analysis, simulation using computer model
Larynx, Vocal Tract (articulators)
Adler, C. H., Bansberg, S. F., Hentz, J. G., Ramig, L. O., Buder, E. H., Witt, K., Edwards, B. W., Krein-Jones, K., & Caviness, J. N., (2004). Botulinum toxin type A for treating voice tremor. Archives of Neurology, 61(9), 1416-1420.
N=13 with VT Video laryngostroboscopy, acoustic analysis
Larynx
Bové, M., Daamen, N., Rosen, C., Wang, C. C., Sulica, L., & Gartner-Schmidt, J. (2006). Development and Validation of the Vocal Tremor Scoring System. The Laryngoscope, 116(9), 1662-1667.
N=10 with VT Transnasal videostroboscopy, acoustic analysis
Larynx
6
speech structures in individuals with vocal tremor, none of these studies
compared the contribution of oscillating speech structures to the final acoustic
output.
To better elucidate the impact of tremor on voice and speech, improved
understanding of the contribution of speech structures on the associated acoustic
signal needs to be prospectively studied. To date, attempts to characterize vocal
tremor by neurogenic disorder has not been successful due to the range of
acoustic patterns demonstrated using primarily fo and SPL rate patterns, in some
cases comparing measures across pitch productions. However, literature
addressing tremor in the limbs has systematically studied the rate, extent, and
conditions under which tremor occurs to classify and diagnose different forms of
tremor. The impact of tremor on functional movements during everyday activities
is considered by neurologists to be the symptoms that bring patients to the clinic.
Similarly, patients with vocal tremor complain of speech and voice problems, but
analysis of the acoustic correlates of the symptoms does not provide insight into
the underpinnings of vocal tremor physiology and its influence on the speech
mechanism. To address the physiologic underpinnings of vocal tremor, a
conceptual model was developed by Barkmeier-Kraemer and Story (2010) (see
Figure 1).
The conceptual model of vocal tremor proposes that tremor oscillation
originating from structures of the respiratory, phonatory, and articulatory systems
will contribute hypothesized patterns of acoustic modulation during voice
production (see Figure 1). For example, tremor oscillations within the respiratory
system are hypothesized to result in acoustic modulation of sound pressure level
7
Figure 1. Barkmeier-Kraemer and Story (2010) Conceptual Model of Vocal Tremor
(SPL) during phonation due to subglottal pressure changes associated with
rhythmic compression and expansion movements of the thoracic cavity due to
tremor affecting muscles of the rib cage, diaphragm, or abdomen (see Figure 1).
Oscillation of the articulatory structures is hypothesized to result in acoustic
modulation of the formant frequencies (i.e., resonant frequencies). The latter is
based upon a model of speech production developed by Brad Story (Story,
1995). This model renders the oral and pharyngeal cavities lined by associated
articulators to behave as a resonating chamber that filters the sound produced at
the level of the larynx by varying length, diameter, and shape. This resonating
chamber is referred to as the vocal tract. If an articulator associated with any
portion of the vocal tract oscillates (e.g., the base of tongue and posterior
Phonatory System Oscillation:
- Vocal fold length changes
o Results in fo
modulation
- Abductory/Adductory vocal
fold movements
o Results in SPL
modulation
Articulatory System
Oscillation:
- Diameter and length
changes in the vocal
tract
- Results in F1 and F2
modulation
Respiratory System
Oscillation:
- Subglottal
Pressure
changes
- Results in SPL
modulation
8
oropharyngeal wall region), the result is oscillation of diameter due to alternation
of constriction and dilation of the vocal tract, or length changes occurring due to
vertical oscillation of the larynx (see Figure 1). Finally, tremor causing oscillation
within the phonatory system is hypothesized to result in two different, or
combined acoustic modulations: 1) lengthwise oscillation of the vocal folds is
hypothesized to predominantly result in modulation of fundamental frequency (fo),
and 2) medial/lateral oscillations of the vocal folds (i.e., oscillation causing
abduction/adduction of the vocal folds) is hypothesized to predominantly result in
modulation of SPL (see Figure 2). Thus, the conceptual model of vocal tremor
offers specific predictions regarding acoustic patterns resulting from tremor
affecting each of the speech mechanism systems to explain the range of vocal
tremor acoustic patterns described in the literature.
The Conceptual Model of Vocal Tremor was developed to help frame
future research investigating characteristics of tremor affecting the speech
Figure 2. Lengthwise (A) and medial/lateral (B) oscillation directions within the Phonatory System
B A
9
mechanism and associated acoustic patterns. For a summary of the model, see
Table 2. To date, testing of this model has primarily been completed using case-
based studies and simulation of tremor within isolated speech mechanism
systems.
One example combined case-based testing of vocal tremor and simulation
of vocal tremor is a study by Lester and colleagues (Lester, et al., 2013). This
study evaluated acoustic patterns in an individual observed to present with
lengthwise vocal fold oscillation during sustained phonation as determined using
stroboscopic imaging. Although lengthwise vocal fold oscillation was
hypothesized within the Conceptual Model of Vocal Tremor to result in a
predominance of acoustic modulation of fo, SPL modulation extent was found to
predominate. Reevaluation of the original stroboscopic evaluation identified that
the laryngeal vestibule appeared to also oscillate in a lengthwise direction
associated with vocal fold lengthwise oscillation. Further investigation was
completed with consideration that the laryngeal vestibule may serve as part of
the vocal tract acting as a resonating chamber and contribute to the acoustic
Table 2. Barkmeier-Kraemer and Story (2010) Summary of Conceptual model of vocal tremor.
Speech Mechanism System Affected by Tremor Hypothesized Acoustic Modulation
Respiratory System oscillation of thoracic cavity compression and expansion
Sound Pressure Level (SPL)
Phonatory System 1: Vocal fold length oscillation Fundamental Frequency (fo)
Phonatory System 2: Medial/lateral vocal fold oscillation
Sound Pressure Level (SPL)
Articulatory System oscillation of diameter and/or length of the vocal tract
Formant Frequencies (F1 and F2)
10
patterns predicted for the vocal tract resulting in formant modulation and
subsequent SPL modulation. Acoustic analysis of the vocal tract formants
supported that the case under study demonstrated formant modulation consistent
with the idea that the laryngeal vestibule contributed to vocal tract oscillation. To
further test the idea of the laryngeal vestibule as a component of the vocal tract
for this individual case, an analysis-by-synthesis approach was utilized. That is,
the acoustic characteristics from the case voice recordings were used to model
acoustic modulation patterns resulting from oscillation originating within the
larynx alone compared to the larynx plus the vocal tract. Systematic analysis of
varied possible conditions of laryngeal and vocal tract oscillation patterns and
associated acoustic patterns demonstrated that the Conceptual Model of Vocal
Tremor helped elucidate the location of tremor within and outside of the vocal
folds of one individual and supported that the laryngeal vestibule resonating
chamber contributed to acoustic modulation patterns as predicted for the vocal
tract (Lester et al., 2013).
Although the case example described above was helpful for testing the
Conceptual Model of Vocal Tremor, general testing on groups of individuals with
vocal tremor has not yet been completed. One reason for this arises due to
difficulty successfully identifying individuals representing tremor isolated within
each system of the speech mechanism. Thus, the question arises whether a
human demonstration of volitional oscillation within one subsystem of the speech
mechanism could be used as a surrogate approach to test the Conceptual Model
of Vocal Tremor (Barkmeier-Kraemer & Story, 2010).
One possible approach to studying vocal tremor acoustic patterns
11
associated with oscillation of speech mechanism structures is to study volitional
modulation of the voice, or vibrato. Western classical singing proponents
consider vibrato to be the “quasi automatic” result of correct singing technique,
used by singers to produce an aesthetically pleasing singing voice (Sundberg,
1994). Vibrato shares many similar acoustic features to vocal tremor such as
rate and extent of fo and SPL (Anand, Shrivastav, Wingate, & Chheda, 2012;
Anand, Wingate, Smith, & Shrivastav, 2012). Similar to vocal tremor, typical
vibrato rate is between 4-7 Hz (Anand, Widgate, et al., 2012; Guzman et al.,
2012; Howes, Callaghan, Davis, Kenny, & Thorpe, 2004; Prame, 1994; Ramig &
Shipp, 1987; Seashore, 1931; Sundberg, 1994; Titze, Story, Smith, & Long,
2002; Watson, Williams, & James, 2012) with an extent of fo modulation between
0.25-2 semitones (Anand, Windgate, et al., 2012; Guzman, et al., 2012; Howes,
et al., 2003; Prame 1994; Seashore 1931). One semitone is equivalent to a
modulating extent of about 6%. Therefore, 0.25-2 semitones would equate to an
extent of about 2-12%. As reported earlier, a typical vocal tremor rate is between
4-8 Hz and is associated with a 3-17% extent of fo.
Also similar to vocal tremor, vibrato is produced by oscillation of structures
within the speech mechanism resulting in acoustic modulation. Based on prior
literature, vibrato studied in trained singers’ results from oscillation in laryngeal
and respiratory structures. Vibrato is predominantly the result of alternating
contraction between the cricothyroid (CT) and a combination of the
thyroarytenoid (TA) and lateral cricoarytenoid (LCA) resulting in laryngeal
oscillations (Dromey & Smith, 2008; Hsiao, Solomon, Luschei, & Titze, 1994;
Sundberg 1994; Titze, et al., 2002). Given the roles of medial/lateral or
12
lengthwise change in the positioning of the vocal folds associated with LCA and
TA/CT contractions, respectively, laryngeal oscillations associated with vibrato
would be expected to result in modulation of SPL and fo acoustic components,
respectively.
Although the predominant involvement of laryngeal musculature appears
associated with TA/LCA and CT musculature, supplementary respiratory (i.e.,
sternocleidomastoid) and postural (e.g., scalenes and latissimus dorsi)
musculature also co-varied activation with production of fo modulation during
production of vibrato (Pettersen & Westgaard, 2005; Watson, et al., 2012).
However, the role of supplementary respiratory musculature was difficult to
interpret from the description in these studies. It is possible that supplementary
respiratory musculature contributes to the artistic aim to achieve balanced
participation between the respiratory and laryngeal systems. For example, as
laryngeal movements associated with production of vibrato occur, SPL changes
can occur due to oscillation in laryngeal valving patterns possibly requiring
supplementary respiratory musculature to respond in an opposite and similar
pattern to achieve stability across speech structures during vibrato performance.
Given that the involvement of the primary expiratory and inspiratory respiratory
musculature was not found, it is likely that the role of supplementary respiratory
musculature was not related to respiratory pressure generation. Therefore,
supplementary respiratory musculature likely serves an antagonistic function
during vibrato generation to maintain stable laryngeal positioning and
performance of the speech mechanism during production of vibrato during
singing.
13
Instruction on the production of vibrato involves extensive training of
techniques that aim to manipulate and balance the use of the respiratory and
laryngeal structures to facilitate modulation of the voice (Kirkpatrick, 2008).
Specifically, vibrato associated with Western classical singing is trained in
singers by teaching the techniques to facilitate oscillation of laryngeal structures
via a reflexive struggle between the CT and TA musculature (Titze et al., 2002).
According to the conceptual model, this would result in lengthwise vocal fold
oscillation resulting in predominant extent of modulation of fo in the acoustic
signal.
Although supplementary respiratory musculature has been implicated
during production of vibrato, respiratory involvement during vibrato is considered
a sign of poor vibrato technique (Kirkpatrick, 2008). However, a healthy use of
the respiratory system for modulation of the voice can be used to facilitate
improved respiratory-phonatory coordination during phonation. One method
documented as successful in achieving this goal is the Accent Method Voice
Therapy (AMVT) (Kotby & Fex, 1998). AMVT involves teaching individuals to
use rhythmic accentuated phoneme productions (Kotby & Fex, 1998) to enhance
phonatory-respiratory coordination during voice production. The rhythmic
accentuation during phonation results from a focus on expiratory rhythmic pulsing
through volitional abdomino-diaphragmatic contractions. Accordingly, these
volitional abdominal accents could be studied to determine whether individuals
can volitionally isolate respiratory oscillation to produce voice modulation
predominantly associated with SPL modulation as predicted by the Conceptual
Model of Vocal Tremor. One study reported a correlation between volitional
14
abdominal accents characteristic of AMVT and a generally associated increase in
SPL and fo (Kotby, Shiromoto, & Hirano, 1993). Although covariation between
SPL and fo was not determined in this study, it did confirm SPL linkage to
volitional abdominal accents. Thus, the AMVT approach to volitional respiratory
system accent production using a rhythmic pattern could be used to test the
Conceptual Model of Vocal Tremor hypothesis of respiratory system contribution
to voice modulation.
Statement of Purpose
Vocal tremor typically presents simultaneously across speech structures
making it difficult to examine predicted contributions of individual portions of the
speech mechanism to predicted acoustic patterns. However, vibrato and AMVT
offer two volitional methods for testing hypothesized contributions of the
laryngeal and respiratory oscillations to acoustic modulation patterns. In
addition, testing these two volitional manipulations of voice production could
elucidate whether the laryngeal and respiratory systems can be isolated from
each other during a voicing task, or are linked in movement patterns. That is,
these voluntary forms of voice modulation (i.e., vibrato and AMVT) could occur
by isolated oscillation of targeted speech structures or may require coordinated
cooscillation of speech structures. If oscillation across systems is demonstrated,
this would support the possibility of volitional motor planning linkages between
speech mechanism structures that might apply to individuals with vocal tremor in
which multiple speech structures appear to be simultaneously affected.
The purpose of this study is to investigate laryngeal and respiratory
15
system physiologic and acoustic correlates as predicted by the Conceptual
Model of Vocal Tremor using trained singers to volitionally produce vibrato and
rhythmic accented production of loudness (i.e., AMVT). The hypothesized
contribution of the respiratory and laryngeal systems to acoustic modulation
patterns are (see Table 3):
1) Rhythmic accented production of loudness using AMVT via chest wall
expiratory movements is hypothesized to be associated with greater
extent of SPL compared to fo modulation.
2) Production of vibrato via lengthwise vocal fold oscillation within the
larynx is hypothesized to be associated with greater extent of fo
compared to SPL modulation.
Table 3. Aims of this study Aims Independent
Variable Dependent Variables
Hypothesized Outcome
1. Study acoustic modulation patterns associated with respiratory system oscillation
Rhythmic accented loudness modulation achieved by AMVT production during sustained phonation
a. Chest wall oscillation rate and extent b. SPL modulation rate and extent c. fo modulation rate and extent
Chest wall movements during accented loudness production will be associated with the rate and extent of SPL > fo modulation
2. Study acoustic modulation patterns associated with laryngeal structure oscillation
Rhythmic fo modulation achieved by vibrato production during sustained phonation
a. Vocal Fold oscillation rate and extent b. fo modulation rate and extent c. SPL modulation rate and extent
Laryngeal oscillation movements during vibrato will be associated with rate and extent of fo > SPL modulation
METHODS
Subjects
This study was approved by the University of Utah Institutional Review
Board (IRB, protocol #IRB 00084972). A total of 11 singers without report of
voicing problems or singing complaints responded to IRB-approved flyers
distributed through social media, campus postings, and email listserves.
Subjects were required to be 40-65 years of age with no less than 5 years as a
trained singer and without report of voicing or singing problems prior to consent
and completion of screening procedures. All consented participants were
screened for the presence of singing and voicing problems (see the description
of the screening procedures below). One individual did not meet inclusion
criteria out of the 11 volunteers for this study. Recruitment continued until a total
of 10 singers meeting inclusion criteria were recruited.
Screening Procedures
All consented participants completed the Voice Handicap Index (VHI)
(Jacobson, et al., 1997) and the Singing Voice Handicap Index (Singing VHI)
(Cohen, et al., 2007) to determine whether voice complaints or singing problems,
respectively, were indicated by abnormal scores (>20 points total score). In
addition, participants completed a questionnaire regarding their singing training
and genre to assure that participants completed a minimum of 5 years of training
17
of the singing voice in Western Classical Singing with a self-identified skill in
producing vibrato and the capability of producing rhythmic accented production of
loudness during sustained phonation (see Appendix a). Of the 10 final
participants, 2 participants initially demonstrated total scores in the abnormal
range on the VHI; however, both individuals clarified that they misunderstood the
VHI rating descriptors and adjusted their scores into the normal range during
discussion regarding the outcomes of the screening procedures.
Procedures
Respiratory kinematic, audio, and laryngeal imaging signals were
recorded simultaneously during sustained phonation tasks. The respiratory
kinematic, audio, and laryngeal imaging signals were synchronized for analysis
of corresponding laryngeal and respiratory kinematic and acoustic patterns. An
example of the simultaneous recording for comparison of laryngeal endoscopy
with simultaneous audio and respiratory kinematic signals can be viewed in
Figure 3.
Respiratory Kinematic Recordings
Respiratory Kinematic Equipment
Two piezo respiratory belt transducers from AD Instruments (Model MLT
1132) were used to measure chest wall movement of the participants. The
transducers were connected to the AD Instruments 8-channel PowerLab (Model
PL 3516) console and LabChart Pro (version 8.1), an AD Instruments software
program. Two piezoelectric respiratory belt transducers were used to measure
respiratory chest wall oscillatory patterns. Each respiband was worn around the
18
Figure 3. LabChart Pro display of simultaneously recorded laryngeal images and acoustic and respiratory signals.
19
participant’s chest wall with one band placed over the rib cage and the
other placed over the abdominal region. The respiratory belt transducer
responsible for measuring rib cage movement was secured around the
circumference of the rib cage at the approximate level of the nipple, running
across the sternum in the front and the upper back. The respiratory belt
transducer responsible for measuring abdominal movement was secured around
the circumference of the abdomen inferior to the ribcage in the front and the
lower back. During the setup, the respiratory belt transducers were positioned to
avoid slipping or repositioning, and to reflect rib cage and abdominal movements
during inhalation, exhalation, and phonation. Also, the recorded signal was
checked to ensure the entire range of chest wall expansion and compression
could be captured within the range of
the LabChart channel.
Respiratory Kinematic Signal Calibration
To calibrate for differences in size and range of utilization of the vital
capacity during singing tasks across participants, the respiratory signal was
normalized across the maximum range of chest wall expansion to compression
during a vital capacity maneuver task. Participants were instructed to inhale air
until they could inhale no further and then to exhale until they could exhale no
further while standing upright. This task was repeated three times. The
maximum recorded value associated with the maximum inhalation maneuver was
set to represent 100% vital capacity, whereas the minimum value associated with
the maximum exhalation maneuver was set to represent 0% vital capacity. The
20
calibration was performed within LabChart and applied to the entire recording for
each participant to enable comparison of respiratory modulation measures within
and between participants.
Respiratory Kinematic Procedures
The signals from the respiratory transducers were recorded in LabChart
(AD Instruments, Version 8.1). Signals were recorded showing the degree of
expansion and compression of the rib cage and abdomen during quiet breathing
and during singing tasks. A summated signal from each portion of the chest wall
was recorded directly onto the AD Instruments LabChart at a sampling rate of 10
kHz. The oscillatory movements of the chest wall were measured for rate and
extent relative to acoustic vibrato and accent (AMVT) patterns.
Acoustic Recordings
Acoustic Recording Equipment
Acoustic recordings were obtained using an AKG head-mounted
condenser microphone (model C520) and preamplifier (Symetrix 302 Dual Mic
Pre-Amp) such that signals were recorded using the AD Instruments 8-channel
PowerLab (Model PL 3516) and LabChart Pro (version 8.1) into the software
simultaneously with laryngeal imaging and respiratory kinematic signals. Audio
signals were recorded at a sampling rate of 40 kHz.
Acoustic Recording Procedures
Once the participant verbally expressed readiness to begin recordings, the
experimental sessions began by providing instruction to the participant to sustain
21
three different vowels, /a/, /u/ and /i/, for the duration of 5 s each. These vowels
were selected to represent “corner vowels” on the English Vowel Chart, produced
using a high forward (i.e., /i/) versus high back (i.e., /u/) and low back (i.e., /a/)
tongue position. It was anticipated that views of the larynx would not be
obstructed by the tongue during production of the two high corner vowels
compared to the occluded views of the larynx during production of the low back
vowel, /a/. The differing vocal production techniques utilized across participants
produced a variety of imaging results, with visibility of the larynx ranging from
fully visible to fully obstructed.
Although 3 corner vowels were recorded, only the recordings of /i/ were
analyzed to test the aims of this thesis. The additional vowel recordings were
conducted as part of a larger data collection for future analysis outside of the
aims of this thesis.
Each participant produced three trials of each vowel with vibrato and
rhythmic accented loudness production using AMVT at comfortable pitch and
loudness. The order of vowel production was not counterbalanced across
participants to assure similar conditions of production for all participants.
However, the order of voice modulation condition for each vowel (vibrato versus
AMVT) was counterbalanced. Three trials for each vowel and voice modulation
condition were produced by each participant. The sequence for production of
each voice modulation condition for each participant was determined in advance
of the recording sessions to assure that equal representation of voice modulation
condition sequencing occurs across all participants. Thus, each participant
produced 18 stimuli (3 vowels: /a/, /u/, and /i/) x 2 voice modulation types (vibrato
22
and AMVT) x 3 trials = 18).
Laryngeal Imaging
Laryngeal Imaging Equipment
Laryngeal imaging was obtained using the Pentax Medical
Nasolaryngoscope System (KayPentax, model 9310HD) and nasoendoscope
(KayPentax, VNL-1070STK) simultaneously with audio and respiratory kinematic
signals via AD Instruments 8-channel PowerLab (Model PL 3516) and LabChart
Pro (version 8.1) hardware/software enabling simultaneous video capture (Figure
3).
Laryngeal Imaging Procedures
Laryngeal imaging was initiated once the audio microphone and
respiratory kinematic bands were placed on the participant. Laryngeal imaging
was obtained using flexible nasoendoscopy to minimize impact on vocal tract
configuration during recording of voicing tasks. The laryngeal imaging
procedures were completed by the supervising thesis advisor (i.e., Dr.
Barkmeier-Kraemer). Topical anesthesia (4% viscous lidocaine) was applied
using a Qtip placed within the left or right entry to the nasal passage and anterior
middle meatus based upon the preference of the subject. The topical anesthesia
was also applied to the scope tip portion of the nasoendoscope posterior to the
lens to minimize discomfort during endoscope placement for laryngeal imaging.
All participants tolerated the scope placement and experimental procedures
without additional need for topical anesthesia.
The nasoendoscope was passed transnasally until the larynx and vocal
23
folds were visible in entirety. Once the nasoendoscope was positioned for
optimal viewing of the larynx during voicing, the participant was instructed to
“warm up” until they felt accustomed to the scope placement during singing.
Once the participant conveyed readiness, the experimental speech tasks were
initiated as described within the audio recording procedures.
DATA COLLECTION AND MEASURES
Classification of Participant Voice Modulation Conditions
by Expert Judges
Three expert judges of the singing voice with at least five years of
experience instructing singers were recruited from the faculty of the University of
Utah School of Music and the School of Musical Theater to complete audio-
perceptual judgments of participant voice recordings and classify each
participant’s recordings as one of the voice modulation types under study.
Hearing Screenings
Each expert judge underwent a hearing screening administered by a
certified SLP to assure typical hearing. The screening tested hearing ability at
500, 1000, 2000, and 4000 Hz at 25 dB SPL in each ear using a pure-tone
audiometer and earphones while sitting in a sound treated booth. The 3 expert
judges included in this study passed the hearing screening according to the
above criteria.
Listening Task Each expert judge individually listened to each participant’s audio
recordings of vibrato and AMVT for each audio recorded trial of /i/. The 3 audio
recording trials of the /i/ vowel in each voice modulation condition were
concatenated into one file for presentation to judges for each of the voice
25
modulation conditions (AMVT vs. vibrato). Each file was presented 3 times for
intrarater reliability purposes. Therefore, 60 total audio samples were presented
to the expert judges (1 vowel x 2 modulation types x 3 presentations of each file
sample x 10 participants = 60 total samples).
These audio files were presented for classification by judges using a
randomized presentation order. Listener stimuli were presented using individual
PowerPoint slides displaying each audio file for evaluation. Judges could
proceed at their own rate through the audio files and replay each file as many
times as necessary to rate the classification of voicing method. Judges listened
to audio file presentations via a high fidelity headset (Sennheiser HD 429) at a
comfortable loudness level while sitting in a sound treated booth. Each listening
trial was classified by each judge as “vibrato,” “accented loudness,” or “unable to
classify.” Judges recorded their ratings for each listening trial onto a formatted
scoring sheet such that the classification of a condition was indicated by circling
the choice that best characterizes what the judge perceived.
Classification of each stimulus was determined by comparing judge
classification ratings to the intended production. A match between judge ratings
and the intended production was scored as an accurate production. A mismatch
between judge ratings and the intended production was scored as an inaccurate
production. Rating scores for the total proportion of participant conditions scored
as accurate were averaged across the three judges’ scores. Recordings from
participants judged as accurate > 75% of the time were used for experimental
analysis. Recordings that did not meet the accuracy criteria above were noted
and reviewed by the authors. A mismatch occurred for the vibrato condition of 2
26
participants and for the accent condition for 1 participant (see Table 4). The
acoustic and respiratory signals from these files were visually inspected and
compared to other recordings judged accurately by judges and did not appear
acoustically dissimilar (see Figure 4A and 4B). Thus, the measures from the
three files judged inconsistently were included within the final data set analyzed
for this study.
Physiologic Measurement of all Recordings
All simultaneously recorded respiratory kinematic, acoustic, and laryngeal
endoscopic signals were saved and stored for each subject recording session.
Two seconds from the midportion of each recorded experimental trial was
selected for analysis. A random number generator was used to code each file
into randomized order to assure blinding of condition, trial, and subject during
measurement of physiologic signals. Nine additional files were randomly
selected (15%) of the total number of files to be analyzed (n = 60) so that
intrarater reliability could be assessed for each physiologic signal measurement
method. Once all measures were completed across all physiologic signals, files
were decoded for statistical analysis. Table 4. Percent agreement between expert judges and intended production condition.
Condition S01 S03 S04 S05 S06 S07 S08 S09 S10 S11
Vibrato 89% *44% *44% 100% 100% 100% 100% 89% 89% 89% AMVT 100% 100% 100% 100% 100% 100% 100% 89% *67%
100%
*This production condition did not meet accuracy criteria and was subject to review by the authors.
27
A.
B.
Figure 4. A comparison of signals rated for each voicing condition (i.e., vibrato and AMVT). A. A comparison of signals rated for the vibrato condition from the two subjects rated with 44% expert agreement from S03 (a-c) and S04 (d-f) and signals with 100% expert agreement from S05 (g-i) versus. B. A comparison of signals rated for the AMVT condition with 67% expert agreement from S10 (a-c) and 100% expert agreement from S05 (d-f).
28
Respiratory Kinematic Analysis
Respiratory oscillation movements were analyzed during the same time
frame analyzed for the audio recordings to compare simultaneous respiratory
kinematic and acoustic patterns.
Respiratory Kinematic Oscillation Rate
The rate of respiratory kinematic oscillation was determined by identifying
peak-to-peak or trough-to-trough patterns of the modulating waveform per
second from the summated rib cage and abdominal voltage signal (see Figure 5).
The rate of oscillation was determined from the number of cycles recorded
per second (Hz).
Respiratory Kinematic Measure Adjustments for Slope
Given that participants sustained phonation while producing vibrato or
AMVT, the lung volume continually decreased across the recorded trials. Thus,
adjustments to the respiratory measures were needed to reduce, or eliminate the
changing lung volume effect on measures obtained over the duration of the
recording. To achieve this, the mean slope of each signal was calculated for
each 2-s segment and factored in to each maximum and minimum value
obtained to adjust for the slope (see Figure 6). The mean slope of the entire
signal was multiplied by the time point associated with the maximum or minimum
value within the 2-s signal and then added to the value measured at the peak
and valley values of the modulating respiratory kinematic signal. Note that the
time point value is based on a 0-2 s time interval, rather than the timestamp of
the entire recording. For example, the 2-s window in Figure 6 took place at
29
Figure 5. Example of determining rate for respiratory kinematic oscillation. Each arrow marks the peak of each modulation cycle displayed in the 2-s window. A total of 11 peak-to-peak cycles are shown in the 2-s window giving a 5.5 Hz respiratory kinematic rate.
Figure 6. Example of measuring extent of respiratory kinematic oscillation. The minimum and maximum values associated with summated rib cage and abdominal movements in the figure represent the original data points. The original data points were corrected before extent was calculated. After adjustment for the sloping values, the relative %VC extent was calculated as described above.
30
timestamp 48-50 s. Therefore, a timestamp of 48.783 s for the data point would
be equal to an adjusted time point value of 0.783 s. The equation used to adjust
the kinematic %vital capacity (%VC) measures to eliminate the slope
effects is shown in equation 1 below: y corrected = y original + [mean slope * time point max/min] (1) Thus, with reference to the values displayed in Figure 6, the following
calculations were completed to obtain the adjusted maximum and minimum
values of one respiratory kinematic cycle:
Cycle 2 minimum corrected = 34.73%VC + [-7.45 * 0.783 s] = 28.90%VC
Cycle 2 maximum corrected = 37.06%VC + [-7.45*0.691 s] = 30.30%VC
Respiratory Kinematic Oscillation Extent
The extent of respiratory kinematic modulation was determined by
measuring the maximum and minimum %VC from the summated rib cage and
abdominal wall voltage signal for each oscillation cycle after the slope adjustment
was applied. The maximum and minimum %VC values were identified by
selecting the peak and trough within the summated rib cage and abdominal wall
signal in LabChart. Then, using the LabChart data pad functions for calculating
maximum and minimum values within a selection, the values were automatically
populated in the data pad spreadsheet. The values were then entered into an
Microsoft Excel spreadsheet for further calculations.
After the original data points were adjusted for the sloping signal, the
%VC extent was calculated by dividing the difference between the minimum
31
%VC value and the maximum %VC value for one cycle by the sum of the
maximum and minimum %VC values for that same cycle and multiplying by 100
to obtain a percentage value. All oscillation cycle relative %VC extent values
were averaged for each trial. Figure 6 can be used again for an example of
respiratory kinematic measurement. Using the adjusted values previously
calculated, the following method was used to determine the relative extent of
respiratory kinematic modulation in equation 2: Extent = (Cycle Max - Cycle Min) / (Cycle Max + Cycle Min) * 100 (2)
Extent of Cycle 2 = (34.15 - 27.85) / (34.15 + 27.85) * 100 = 10.2%
Acoustic Measures
Acoustic modulation patterns of fo and SPL were measured from the
middle 2-s portion of each recorded trial. The fo and SPL values within each
selected acoustic segment for analysis were displayed in Praat (Boersma &
Weenink, 2015; v 5.4.09) for analysis of rate and extent of modulation.
Acoustic Modulation Rate
The 2-s segments of fo and SPL modulation were analyzed for rate by
counting the number of peak-to-peak or trough-to-trough fo and SPL modulation
cycles displayed and dividing by 2-s to record rate of modulation in Hertz
(cycles/s) (see Figure 7). The number of fo and SPL modulation cycles per
s was determined for the signals displayed in Figure 7 with the following equation 3:
32
Figure 7. Example of fo (top line) and SPL (bottom line) plot of vibrato within Praat. Arrows have been added to the signal to indicate the beginning of the cycle. Rate (Hz) = (Number of cycles) / (time (s)) (3)
Rate for fo (top signal) is 9.5 cycles / 2-s = 4.8 Hz
Rate for SPL (bottom signal) is 9.5 cycles / 2-s = 4.8 Hz
Acoustic Modulation Extent
The extent of fo and SPL modulation for each cycle was determined by
measuring the maximum and the minimum values from the peak and valley
portions of each cycle (see Figure 8). The maximum and minimum values were
identified by importing the 2-s portion of the acoustic files being analyzed into
Praat. The modulation cycles were identified using the peak to peak or trough to
trough analysis (see Figure 8). Each cycle was then highlighted by dragging the
cursor across one cycle. Then, functions were completed within Praat to
calculate the maximum and minimum values of fo and SPL (i.e., get minimum
pitch, get maximum pitch, get minimum intensity, and get maximum intensity).
The respective values were then entered into an MS Excel spreadsheet for
calculation of the extent values.
33
Figure 8. Example of fo (top line) and SPL (bottom line) plot of vibrato within Praat. Arrows have been added to the signal to indicate the maximum and minimum points of cycle 3 (fo) and cycle 8 (SPL). fo Extent Measures
Calculation of fo extent was completed by subtracting the minimum fo value
from the maximum fo value and dividing the resulting value by the sum of the
maximum and minimum fo values of that cycle and multiplying by 100 to
determine a percentage value. This was repeated for all fo modulation cycles
within the selected segment of each trial as shown in the equation 4 below for Figure 8: Extent = (fo Max - fo Min) / (fo Max + fo Min) * 100 (4)
Figure 8 Extent fo = (246 Hz - 237 Hz) / (246 Hz + 237 Hz) * 100 = 1.9%
SPL Extent Measures
The extent of SPL modulation was determined from the maximum and
minimum dB SPL values associated with SPL modulation (see Figure 8). SPL
values were converted to a linear scale, Pascals, so that the same calculation
method used for fo extent could be performed for SPL extent. Using the values
shown in Figure 8 below, the following equation 5 offers an example of one cycle
of SPL modulation extent calculation:
34
SPL Extent = (SPL Max – SPL Min) / (SPL Max + SPL Min) * 100 (5)
Figure 8 Extent SPL = (.017 Pa - .016 Pa) / (.017 Pa + .016 Pa) * 100 = 3.03%
Laryngeal Imaging Kinematic Analysis
Frame by frame analysis of laryngeal oscillatory movements was
completed to determine the predominant pattern of laryngeal movement
associated with each voice modulation condition (vibrato vs. AMVT). Laryngeal
oscillation movements were analyzed from a similar or the same 2-s duration
segment as the respiratory kinematic and acoustic signals for each experimental
trial. Adjustments in the portion of the endoscopic recording were made if
pharyngeal or laryngeal postures occluded views of the vocal folds. During the
latter situations, the 2-s segment analyzed was shifted earlier or later as needed
to assure continuous views of the vocal folds for kinematic analysis. Given the
subjective impression that laryngeal movements remained continuous during all
trials, it was not expected that analysis from a shifted time segment would impact
measurement outcomes. The maximum and minimum range of vocal fold
movement in the anterior/posterior (i.e., lengthwise) and medial/lateral (i.e.,
abductor/adductor) directions was measured to determine laryngeal oscillation
rate and extent for comparison to acoustic and respiratory kinematic modulation
patterns. The endoscopic video recordings for each subject were imported into
QuickTime to select each of the mid-portion 2-s segments analyzed for the
acoustic and respiratory kinematic signals. The maximum and minimum
movement frames associated with each laryngeal movement cycle within the
analyzed segment were copied and imported into ImageJ (October, 2015) for
35
measurement.
Laryngeal Oscillation Rate
Laryngeal oscillation rate was determined by recording the number of
laryngeal oscillatory cycles obtained during the 2-s segment analyzed for
each trial divided by 2 s to determine the rate in Hertz (cycles/s).
Laryngeal Oscillation Extent
Extent of laryngeal oscillation was measured by identifying image frames
displaying the minimum and maximum displacement of laryngeal structures/vocal
folds during the predominant direction of oscillation (i.e., lengthwise versus
abduction/adduction). The lens to larynx distance varied during recordings,
requiring image measures to be normalized by creating a ratio of distance (unit =
pixels) between laryngeal movement measures and an anatomical measure
observed to remain constant (e.g., width of the epiglottis apex). Referent Anatomical Distance Measures
Individual frames judged to exhibit the maximum displacement end points
of laryngeal oscillation were selected and saved for analysis within ImageJ
software (Rasband, 2015). The referent anatomical distance between two
constant and clearly identifiable locations measuring the interarytenoid distance
was measured for each maximum and minimum cyclic displacement image (see
Figure 9, Part A and B). This was accomplished by drawing a line between the
two referent image points using the ImageJ line tool and recording the length of
the line in pixels.
36
Figure 9. An example of laryngeal imaging measures for the first end point of one laryngeal oscillatory cycle. Each panel displays A) the image analyzed, B) the anatomical referent line measure, C) the relative measure of vocal fold length, and D) the relative measure of interarytenoid distance associated with the first end point of one laryngeal oscillation cycle. B. Example of laryngeal imaging measures for the second end point of one laryngeal oscillatory cycle. Each panel displays A) the image analyzed, B) the anatomical referent line measure, C) the relative measure of vocal fold length, and D) the relative measure of interarytenoid distance associated with the second end point of one laryngeal oscillation cycle.
37
A.
B.
38
Laryngeal Lengthwise Extent Measures
The length of the vocal folds during maximum and minimum laryngeal
cyclic movements was measured by drawing a line from the anterior commissure
(or most anterior visible portion of the vocal folds) and posterior commissure (or
most visible posterior point on the vocal folds) and recording the number of pixels
associated with the length. The magnitude, or extent, of lengthwise vocal fold
movement during the laryngeal movement cycle associated with vibrato or AMVT
was determined by dividing the vocal fold length measure (in pixels) by the
standard referent distance (in pixels) (see Figure 9). The extent of lengthwise
change was determined using the following equation (6) for
each cycle of maximum and minimum movements:
% vocal fold lengthwise extent = (vocal fold maximum length – vocal fold
minimum length) / (vocal fold maximum length + vocal fold minimum
length) * 100 (6)
In the laryngeal oscillation cycle shown in Figure 9, the extent of laryngeal abduction/adduction extent would be: Relative VF Lengthwise Extent = (.25 - .24) / (.25+.24) * 100 = 2.0% Laryngeal Abduction/Adduction Extent Measures
The interarytenoid distance was measured during maximum and minimum
laryngeal cyclic movements to determine changes in abduction and adduction of
the vocal folds. Due to the absence, or small number of pixels associated with
the glottal width between vocal processes, the interarytenoid distance was used
39
as a surrogate measure to assure adequate sampling distance for analysis. In
the absence of a glottis during full approximation of the vocal folds, the measure
would yield 0 pixels and would cause errors in the calculation of extent. Thus,
interarytenoid distance was determined to reflect changes in glottal width for the
purposes of this study.
The interarytenoid distance was measured by drawing a line between the
most lateral visible point on the superior surface of the left vocal process and the
opposite lateral visible point on the right vocal process (see Figure 9, Part A and
B, Image D). The extent of abduction/adduction change was determined using
the following equation (7) for each cycle of maximum and minimum movements:
% vocal fold ABDuction / ADDuction extent = (vocal fold maximum
abduction – vocal fold minimum abduction) / (vocal fold maximum
abduction + vocal fold minimum abduction) * 100 (7)
In the laryngeal oscillation cycle shown in Figure 9, the extent of laryngeal abduction/adduction extent would be:
Relative Vocal Fold Abductor/Adductor Extent =
(.14 - .13) / (.14 + .13) * 100 = 3.7%
Statistical Analysis
The dependent variables for investigation are the average rate and extent
of fo and SPL acquired from acoustic recordings; the average rate and extent of
respiratory kinematic signals acquired from movements of the chest wall; and the
average rate and extent of laryngeal kinematic images acquired from lengthwise
40
laryngeal movements and abductor/adductor laryngeal movements. Each
variable was compared within and across voice modulation conditions, vibrato
and AMVT (see Table 5).
Statistical analysis was completed to determine whether hypothesized
differences between the extent of SPL and fo for each location of oscillation
(respiratory system versus laryngeal system) occurred. To accomplish this, a
mixed effects logistic regression was completed to compare dependent variable
outcomes predicted between respiratory kinematic measures and associated
acoustic modulation measures (%SPL extent > % fo extent) and laryngeal
imaging kinematic measures and associated acoustic modulation measures
(lengthwise oscillation = % fo extent > %SPL extent; abductor/adductor oscillation
= %SPL extent > % fo extent).
Intrarater reliability was also evaluated using an Intra Class Correlation
(ICC) approach including calculation of 95% confidence intervals on the 9 files
Table 5. Intrarater reliability determined using intraclass correlations.
Dependent Variable AMVT (N=3) Vibrato (N = 6)
Avg fo Rate (Hz) 0.882 (-0.001 ~0.997) 1 (NaN ~ NaN)
Avg fo Extent (%) 0.994 (0.913 ~1) 0.999 (0.996 ~ 1)
Avg SPL Rate (Hz) 0.857 (-0.105 ~0.996) 0.878 (0,44 ~0.982)
Avg SPL Extent (%) 0.807 (-0.262 ~0.995) 0.986 (0.922 ~0.998)
Avg Respiratory Kinematic Rate (Hz) 1 (NaN ~NaN) 1 (.61 ~ 1)
Avg Respiratory Extent (%) 1 (NaN ~NaN) 1 (.61 ~ 1)
Avg Laryngeal Rate (Hz) 0.977 (0.683 ~ 0.999) 0.723 (0.02 ~ 0.955)
Avg vocal fold length Extent (%) 0.934 (.293 ~0.998) 0.925 (0.622 ~ 0.989)
Avg vocal fold abduction/adduction extent (%) .489 (-0.692 ~0.983) 0.723 (0.02 ~0.955)
41
randomly selected for repeated measures. The 9 measures were only used for reliability measures and were not included in the final analysis.
Intrarater Reliability
Intraclass Correlation (ICC) results demonstrated high levels of intrarater
reliability levels for all dependent variables except for the average vocal fold
abduction/adduction extent (%) measure under the AMVT condition which
achieved moderate reliability (i.e., 0.5 – 0.6) (see Table 5). The lower reliability
levels for vocal fold abduction/adduction measures under the AMVT condition
may relate to the lower number of files analyzed (n = 3), or may reflect the
greater difficulty determining laryngeal oscillation patterns given significantly
slower rates of oscillation that included vertical laryngeal movements.
RESULTS
Qualitative Analysis of Results
Tables 6 and 7 provide a summary of the average measures and their
standard deviations for each subject under the AVMT and vibrato conditions,
respectively. As can be seen, individual singers generally exhibited similar rates
of acoustic modulation within each of the singing conditions (AMVT versus
vibrato). That is, the AMVT condition was associated with slower acoustic
modulation rates for both fo and SPL (i.e., approximately 2-4 Hz) than for the
vibrato condition (i.e., approximately 5-7 Hz). Similarly, laryngeal kinematic rates
for vocal fold length change and abduction and adduction of the vocal folds were
slower under the AMVT voicing condition than for the vibrato condition.
Respiratory kinematic rates were absent under the vibrato condition and present
at a similar rate to laryngeal kinematic rates for AMVT.
Respiratory System Contributions to Acoustic Modulation
It was hypothesized that the AMVT condition would be associated with a
predominant oscillation of the chest wall contributing to a greater extent of SPL
acoustic modulation compared to the vibrato condition. As shown in Tables 6
and 7, the average respiratory kinematic extent for all subjects during the AVMT
condition was measured at 47.5% (SD = 1.2%). For the vibrato condition,
respiratory kinematic extent was measured at 0% (SD = 0%). The respiratory
43
Table 6. AVMT voicing condition descriptive statistical summary for individual subjects.
*Signifies measures that are significantly different between voicing conditions. Table 7. Vibrato voicing condition descriptive statistical summary for individual subjects.
*Signifies measures that are significantly different between voicing conditions.
1 3 4 5 6 7 8 9 10 11
Avg fo Rate (Hz)
(SD) 4.5 (0) 5 (0) 6.2 (0.3) 5 (0) 5.2 (0.3) 5.3 (0.3) 6 (0) 4.5 (0) 5.3 (0.3) 4 (0) 5.1 (0.7)*
Avg fo Extent (%)
(SD) 3.5 (0.2) 4.8 (0.4) 4.1 (0.7) 4.4 (0.3) 2.2 (0.2) 3.7 (0.4) 1.9 (0.1) 4.7 (0.9) 3.4 (0.3) 2.0 (0.4) 3.5 (1.1)
Avg SPL Rate (Hz)
(SD) 4.5 (0) 4.8 (0.3) 6.2 (0.3) 5 (0) 5 (0.5) 5.2 (0.3) 5.8 (0.3) 4.5 (0) 5.2 (0.3) 4 (0) 5 (0.6)*
Avg SPL Extent
(%) (SD) 4.5 (1.6) 8.9 (0.4) 19.6 (5.7) 7.2 (1.0) 2.8 (0.5) 6.9 (1.3) 6.7 (0.3) 9.1 (3.5) 12.1 (0.7) 4.4 (1.4) 10.0 (0)*
Avg Respiratory
Kinematic Rate
(Hz) (SD)
0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0)*
Avg Respiratory
Kinematic Extent
(%) (SD)
0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0)*
Avg Laryngeal
Kinematic Rate
(Hz) (SD)
4.2 (1.9) 6 (1.3) 10.2 (2) 6.3 (0.8) 6.5 (3) 7.5 (0) 10.3 (1.6) 6.5 (0.5) 6.3 (1.6) 4.8 (2.5) 6.9 (2)*
Avg Vocal Fold
Lengthwise
Extent (%) (SD)
1.8 (1.2) 2 (0.2) 3.3 (1.8) 6.3 (1.3) 0.9 (0.1) 5.4 (1.9) 1.5 (0.4) 3.8 (0.5) 3.7 (0.7) 2.9 (1.1) 3.2 (1.7)
Avg Vocal Fold
Abduction/Adduct
ion Extent (%) (SD)
3.9 (0.8) 1.5 (0.2) 4.7 (0.3) 4.8 (1.7) 2.2 (0.9) 5.9 (1.9) 3 (0.6) 6.6 (1.2) 4.1 (0.1) 2.8 (0.7) 3.9 (1.6)
SUBJECTSMEASURES
ACOUSTIC
RESPIRATORY KINEMATICS
LARYNGEAL KINEMATICS
ALL
SUBJECTS
1 3 4 5 6 7 8 9 10 11
Avg fo Rate (Hz)
(SD) 3.8 (0.8) 2.5 (0) 2.2 (0.3) 3.7 (1) 3.5 (1.3) 2 (0.9) 2.2 (0.3) 3.5 (0.9) 2.5 (0.5) 1.7 (1.2) 2.8 (0.8)*
Avg fo Extent (%)
(SD) 6.1 (3.2) 8.9 (0.2) 2.8 (1.0) 1.6 (0.3) 2.7 (0.9) 2.8 (0.2) 1.7 (0.2) 4.2 (0.4) 2.6 (0.5) 2.0 (0.5) 3.5 (2.3)
Avg SPL Rate (Hz)
(SD) 3.5 (0.5) 2.5 (0) 2.3 (0.3) 1.8 (0.3) 1.5 (0) 1.5 (0) 2 (0) 3 (0) 2 (0) 1 (0) 2.1 (0.7)*
Avg SPL Extent
(%) (SD)
28.4
(13.1)
49.5
(5.4)35 (8.9)
12.6
(6.7)
32.6
(10.6)
68.2
(08.7)23 (1.3)
49.2
(1.9)
28.3
(1.6)
37.7
(9.9)40 (20)*
Avg Respiratory
Kinematic Rate
(Hz) (SD)
3.7 (0.3) 2.5 (0.5) 2.7 (0.8) 2 (0) 1.7 (0.3) 1.7 (0.3) 2 (0) 3 (0) 2.2 (0.3) 1.8 (0.6) 2.3 (0.6)*
Avg Respiratory
Kinematic Extent
(%) (SD)
46.2
(1.2)
48.5
(0.4)45.7 (1.4)
48.5
(1.3)47 (3)
48.6
(0.1)47.6 (1.1)
48.4
(1.1)
48.6
(0.5)
46.1
(4.7)47.5 (1.2)*
Avg Laryngeal
Kinematic Rate
(Hz) (SD)
6.7 (1.8) 4.2 (0.3) 2.5 (1) 3.7 (3.2) 2.3 (1.3) 2.7 (0.3) 1.8 (1.6) 5.2 (1) 3.3 (0.8) 1.5 (1.5) 3.4 (1.6)*
Avg Vocal Fold
Lengthwise
Extent (%) (SD)
1.8 (0.4) 2.8 (0.5) 2.4 (0.5) 3.1 (2.7) 2.2 (1.5) 4.3 (1.5) 0.9 (1.6) 3 (1.7)10.7
(3.2)0.7 (0.8) 3.2 (2.8)
Avg Vocal Fold
Abduction/Adduct
ion Extent (%) (SD)
3.4 (1.5) 3.3 (1.3) 4.9 (1.7) 3 (2.6) 3 (1.3) 3.3 (0.7) 1.4 (2.5) 7.6 (1.5) 3.7 (1.6) 3 (3) 3.7 (1.6)
LARYNGEAL KINEMATICS
MEASURESSUBJECTS ALL
SUBJECTS
ACOUSTIC
RESPIRATORY KINEMATICS
44
kinematic extent was demonstrated to be significantly greater, on average, under
the AMVT condition than for vibrato (p < .001). Refer to Figure 10 for a graphical
comparison of respiratory kinematic extent across the voicing conditions (i.e.,
AMVT and vibrato).
The average SPL modulation extent for the AVMT condition was
measured at 40% (SD=20%), whereas the average SPL extent for the vibrato
condition was 10% (SD=0%). The acoustic measure of SPL extent was also
significantly greater under the AMVT condition, on average, than for vibrato (p =
.026) supporting the hypothesized contribution of the respiratory system to voice
modulation. Interestingly, the rate of SPL modulation was also found to
significantly differ between voicing conditions. Respiratory oscillation under the
AMVT condition was associated with significantly slower rate of SPL modulation
than for the vibrato condition (p < .001). Refer to Figures 11, 12, and 13 for a
graphical comparison of kinematic oscillation rates, SPL modulation extent, and
SPL modulation rate, respectively, across voicing conditions (i.e., AMVT and
vibrato).
Phonatory System Contributions to Acoustic Modulation
It was hypothesized that the vibrato condition would be associated with a
predominant laryngeal kinematic oscillation, vocal fold lengthwise oscillation,
resulting in greater fo modulation extent compared to the AVMT condition. As
shown in Tables 6 and 7, The average vocal fold length extent for the vibrato
condition was measured at 3.2% (SD = 1.7%). The average vocal fold length
extent for the AVMT condition was measured at 3.6% (SD = 3%). Statistical
45
Figure 10. Kinematic extent comparisons between voicing conditions (i.e., AMVT
and vibrato)
Figure 11. Kinematic extent comparisons between voicing types (i.e., AMVT and vibrato)
46
Figure 12. Acoustic extent comparison between voicing conditions (i.e., AMVT and vibrato)
Figure 13. Acoustic rate comparison between voicing conditions (i.e., AMVT and vibrato)
47
evaluation showed that the laryngeal kinematic extent was not significantly
greater under the vibrato condition. Rather, the laryngeal kinematic extent did not
significantly differ between the two voice modulation conditions (p = .95). Refer
to Figure 13 for a graphical comparison of laryngeal kinematic extent across the
voicing conditions (i.e., AMVT and vibrato).
The average fo modulation extent for the vibrato condition was measured
at 3.5% (SD = 1.1%). The average fo modulation extent for the AVMT was 3.5%
(SD = 2.3%). The acoustic measure of fo extent also did not differ significantly
under the two voice modulation conditions (p = .92) as was hypothesized.
However, the rate of fo modulation was found to significantly differ between
voicing conditions demonstrating that laryngeal oscillation under the vibrato
condition was a significantly faster rate than for the AVMT condition (p < .001).
Furthermore, the average laryngeal kinematic rate was significantly higher for the
vibrato condition compared to the AVMT condition (p < .001). Refer to Figures 12
and 13 for a graphical comparison of fo modulation extent and rate, respectively,
across voicing conditions (i.e., AMVT and vibrato).
Vocal Tract Movements by Condition
Vertical Laryngeal Movement The presence or absence of vertical laryngeal movement was recorded
during laryngeal kinematic analysis and observed to demonstrate predominant
patterns associated with each voicing condition. During the AVMT condition, 70%
of participants demonstrated vertical laryngeal movement in at least 2 of 3 trials
or more (67%). In contrast, only 40% of participants demonstrated vertical
48
laryngeal movement in at least 2 of 3 trials in the vibrato condition.
Pharyngeal Movement
The presence or absence of pharyngeal constriction was recorded during
laryngeal kinematic analysis for comparison between voicing conditions. During
the AVMT condition, 30% of participants demonstrated pharyngeal movement in
at least 2 of 3 trials compared to 60% of participants during the vibrato condition.
DISCUSSION
The purpose of this study was to test the contribution of respiratory and
laryngeal oscillation patterns to acoustic modulation patterns. The conceptual
model of vocal tremor developed by Barkmeier-Kraemer and Story (2010)
proposed that oscillation of the respiratory system would be reflected by SPL
modulation patterns in the acoustic signal. They also proposed that laryngeal
oscillation patterns would be associated with acoustic patterns specific to the
laryngeal kinematic patterns exhibited. Vocal fold length oscillation was
hypothesized to cause fo extent modulation whereas vocal fold abduction/
adduction oscillation would affect interarytenoid distance and be reflected in
acoustic SPL extent modulation. The current study investigated voluntary
manipulation of laryngeal and respiratory oscillation within trained singers to
study associated patterns of acoustic modulation. This was achieved by
comparing AVMT and vibrato voicing methods during sustained phonation.
AVMT modulates the voice using accented contraction of the respiratory system
(Kotby & Fex, 1998). In contrast, vibrato utilizes vocal fold lengthwise oscillation
to modulate fo (Dromey & Smith, 2008; Hsiao, Solomon, Luschei, & Titze, 1994;
Sundberg 1994; Titze, et al., 2002).
50
Respiratory System Contributions to Acoustic Modulation
The results of this study supported the hypothesized contribution of
respiratory oscillation to acoustic modulation. The AVMT voicing condition was
used to provide respiratory system oscillation during voicing modulation and was
shown to predominantly be associated with greater respiratory kinematic
oscillation extent and SPL modulation extent. Kotby, Shiromoto, & Hirano (1993),
reported that significant SPL modulation occurred associated with abdomino-
diaphragmatic contraction during production of the AVMT method of voicing. The
current study’s outcomes lend support to physiologic models of phonation
proposing modulation of SPL associated with respiratory system compression
and expansion movements during voicing (Barkmeier-Kraemer & Story, 2010;
Farinella, Hixon, Hoit, Story, Jones, 2006; Story, 1995;). The distinct difference
between respiratory kinematics and acoustic modulation patterns during the
AMVT and vibrato conditions supports prior literature showing that the chest wall
portion of the respiratory system does not appear to contribute toward natural
vibrato production in trained singers (Pettersen & Westgaard, 2005; Watson, et
al., 2012).
The slow respiratory oscillation rate measured during the AVMT condition
distinguished respiratory oscillation from laryngeal oscillation in this study. The
slower rate of oscillation and acoustic SPL modulation during the AMVT condition
compared to the vibrato condition may relate to the larger mass of the respiratory
structures and musculature such as the diaphragm, abdominal muscles, and
intercostal muscles. In contrast, the laryngeal skeletal framework, structures,
and musculature are significantly less massive likely enabling the faster
51
oscillation rates measured for the larynx (Dalvi & Premkumar, 2011).
The findings revolving around the AVMT voicing condition have important
clinical implications for clinical evaluation and management of vocal tremor.
Individuals with symptoms of vocal tremor exhibiting greater acoustic SPL
modulation extent than fo modulation extent at a rate closer to 3 Hz should be
evaluated for respiratory contributions to their vocal tremor. In cases where the
respiratory system is the predominant contributor to vocal tremor, consideration
regarding optimal medical or behavioral management would be required. For
example, vocal tremor is most commonly medically treated by injecting Botox®
into the intrinsic laryngeal musculature (i.e., thyroarytenoid, interarytenoid, or
posterior cricoarytenoid muscles) (Kendall & Leonard, 1995; Schneider &
Deuschl, 2015). This is warranted if the vocal tremor is predominantly caused by
tremor within the intrinsic laryngeal muscles. However, vocal tremor caused by
the respiratory system and not the laryngeal musculature may not provide
optimal results using Botox® injections into the laryngeal musculature (Bove et
al., 2006). In addition, behavioral clinical treatment typically involves increased
recruitment of respiratory contraction during phonation to offload laryngeal and
throat muscle tension. Such a speech treatment approach may be more difficult
for individuals to perform if the source of their vocal tremor is from the respiratory
system.
52
Laryngeal Contributions to Acoustic Modulation
Laryngeal oscillation contributions to vocal tremor were hypothesized to
result in greater fo modulation extent acoustically. This was tested in the current
study by comparing the correspondence between vocal fold lengthwise and
abduction/adduction oscillations during vibrato to modulation of fo extent. The
hypothesized differences between acoustic measures of fo extent and vocal fold
kinematics during AVMT and vibrato were not supported. In contrast, vibrato and
AVMT voicing conditions were shown to equally contribute to fo extent modulation
and vocal fold lengthwise and abduction/adduction laryngeal oscillation. Although
laryngeal kinematic patterns appeared similar between the AMVT and vibrato
conditions, the kinematic rates were significantly different between voicing
conditions and closely corresponded with the rate of acoustic modulation for
each. That is, the rate of laryngeal kinematic patterns was significantly faster
during the vibrato condition than for AMVT. These findings suggest that the
laryngeal kinematic patterns measured during the AMVT condition were largely
influenced by respiratory system kinematics. This is not entirely surprising given
that the larynx is considered to be part of the respiratory system. The utilization
of the larynx for voice production cannot be disassociated from its reliance on
respiratory pressure and flow patterns for which laryngeal configuration may
adjust associated with lung volume levels (Lowell, Barkmeier-Kraemer, Hoit,
Story, 2008).
The determination that laryngeal kinematics are similar, but slower during
AMVT compared to vibrato demonstrates the reliance of laryngeal configurations
on respiratory functions during voice production. In addition, the finding that
53
abduction/adduction vocal fold movements were a component of vibrato is in
contrast to prior literature reporting vibrato production associated with the
antagonist relationship of the TA and CT muscles (Dromey & Smith, 2008; Hsiao,
Solomon, Luschei, & Titze, 1994; Sundberg 1994; and Titze, et al., 2002). The
findings in this study suggest that vibrato is produced using a complex interaction
between the phonatory, respiratory, and articulatory (i.e., vocal tract) systems.
Another interesting finding of this study was the observation of inferior
constrictor muscle activation associated with laryngeal movements during
vibrato. This was also not entirely surprising to observe given the suspension of
the larynx within the throat by anterior and posterior throat musculature. Thus,
observation of counter-contraction of the thyropharyngeus musculature opposite
laryngeal vibrato movements is speculated to demonstrate postural stabilization
via pharyngeal constrictor muscle activation.
The influence of the respiratory condition in this study (i.e., AVMT) on
laryngeal kinematic patterns suggests that laryngeal behaviors are not
independent from other speech mechanism structures. This is consistent with
prior findings in individuals diagnosed with vocal tremor. For example, a prior
study by Lester et al. (2013) tested the contribution of lengthwise vocal fold
oscillation to acoustic modulation patterns and did not find support for the
predicted contribution of laryngeal oscillation to fo extent modulation. However,
upon expanded acoustic analysis, implication of the vocal tract was found as
indicated by formant modulation. Upon further review of the laryngeal imaging
recordings, it was discovered that the epilarynx, or laryngeal vestibule oscillatory
patterns, were likely affecting formant locations within the vocal tract and affected
54
SPL extent modulation as predicted by the conceptual framework model. In this
study, systematic analysis of laryngeal vocal fold movements in addition to
presence/absence of pharyngeal constriction movements documented that
laryngeal movements during vibrato (i.e., predominantly laryngeal involvement)
were also frequently associated with pharyngeal constriction movements in
addition to laryngeal vestibule compression and expansion. These findings
suggest that laryngeal oscillations are likely to contribute to alteration of vocal
tract configuration requiring co-contraction of vocal tract musculature involved in
laryngeal posturing functions. Additional investigation implementing
electromyography (EMG) to improve upon prior literature investigating the
involvement of supplementary respiratory musculature would help resolve the
timing and role of observed vocal tract structure involvement during vibrato
production (Finnegan et al., 2003; Tomoda et al., 1987).
In general, the outcomes of this study demonstrate that the vibrato
condition shows close correspondence between laryngeal kinematic patterns and
respiratory system behaviors. These findings support that speech treatment for
vocal tremor that trains improved utilization of the respiratory system may
influence laryngeal configurations and muscle contractions associated with
laryngeal-based vocal tremor. Importantly, the findings of this study demonstrate
significant differences in the rate of vocal tremor associated with laryngeal versus
respiratory sources of oscillation as well as acoustic SLP extent modulation
patterns. These findings support the proposal of the conceptual model of vocal
tremor (Barkmeier-Kraemer and Story, 2010) that the speech structures affected
by tremor may be identified by their contributions to the acoustic modulation
55
patterns. However, hypothesized contributions of the larynx to acoustic fo extent
associated with vocal fold lengthwise oscillation was not successfully tested due
to linkage between respiratory and laryngeal behaviors during vibrato and AVMT
voicing.
Application of Current Findings to Vocal Tremor
The results of this study offer further support for the respiratory
contributions to voice modulation as hypothesized by the conceptual model of
vocal tremor (Barkmeier-Kraemer & Story, 2010). The current findings showed
correspondence between respiratory kinematic patterns and SPL rate and extent
of modulation within the AMVT condition. Although respiratory contributions to
vocal tremor have not been directly studied, the presence of respiratory structure
oscillation in individuals with vocal tremor has been reported in the literature
(Tomoda et al., 1987). Future work will need to address similar patterns in those
with vocal tremor to confirm similar association of respiratory kinematics and
voice modulation.
One problem that has impeded testing of hypothesized laryngeal
contributions to voice modulation in vocal tremor is the cooccurrence of
oscillation of vocal tract structures with laryngeal oscillations. The current study
aimed to isolate laryngeal oscillation through the implementation of vibrato
compared to AMVT in trained singers. We hypothesized that laryngeal oscillation
would be absent during production of AMVT compared to vibrato. However, the
trained singers in this study did not demonstrate differences in vocal fold
lengthwise and abduction/adduction kinematic patterns between the AMVT and
56
vibrato conditions as hypothesized. The primary distinction between laryngeal
kinematic oscillation patterns in the AMVT and vibrato conditions was the rate of
movement. The rate of laryngeal kinematic patterns was significantly slower
during the AMVT condition than during the vibrato condition. That is, kinematic
pattern rate distinguished between laryngeal oscillation conditions rather than
specific SPL or fo extent of modulation. As such, future work investigating
laryngeal kinematic patterns associated with acoustic modulation in those with
vocal tremor may benefit from a comparison of acoustic rate patterns to
determine the source of oscillation within the speech mechanism. Thus, future
research needs to compare and contrast all three sources of tremor (i.e.,
respiratory, laryngeal, and vocal tract) to further test and refine the conceptual
model of vocal tremor.
Limitations
The current study offered important contributions toward understanding
the contributions of laryngeal and respiratory oscillation to SPL and fo acoustic
patterns. Future research on this topic could improve upon the current findings
by consideration of methodology limitations of this study.
Laryngeal kinematic measures were limited by the use of nasoendoscopy
to analyze the dynamic larynx without a calibrated light grid, or other
measurement calibration methods that would have improved upon the accuracy
of this study’s kinematic measures. That said, the reliability of the laryngeal
kinematic measures was highly reliable for all but the abduction/adduction
measures specific to the AMVT condition. However, the range of findings for the
57
latter was likely due to the smaller randomly chosen files for analysis that were at
least moderately reliable. Future research could be improved by the use of
electromyography to study musculature associations with observed movements
and incorporation of calibration lighting as this technology improves.
The within-subjects design of this study did not require calibration of SPL
for comparison within individuals between two conditions of voicing. SPL
calibration signals were recorded; however, the calibration factors caused peak
clipping of recorded signals due to the large amplitude signals recorded by
singers in this study. Thus, the decision was made to utilize relative SPL. Thus,
relative SPL values were recorded and compared within subjects between the
two conditions. However, reporting of SPL values in this study requires caution
as averaged measures were relative SPL. In future work, calibrated SPL values
would enable comparison of SPL between individuals.
Another future consideration would be to include aerodynamic measures
of SPL and airflow. Similarly, consideration of laryngeal behaviors associated
with lung volume levels would help interpret linked laryngeal and respiratory
behaviors during vibrato and AVMT conditions. It is possible that laryngeal
oscillation patterns vary between voicing initiation at higher lung volumes
compared to voicing toward the end of voicing (Lowell et al., 2008). Such factors
may be important to study associated with structural kinematic behaviors to
elucidate further the role of the larynx relative to respiratory and vocal tract voice
production conditions.
Finally, generalization of the findings of this study to the larger population
of singers will require increased numbers of participants in future work. However,
58
the robust outcomes of this study would be expected to be supported in future
larger studies.
Conclusion
The purpose of this study was to test the conceptual model of vocal tremor
developed by Barkmeier-Kraemer and Story, 2010 by linking respiratory and
laryngeal kinematic oscillations to acoustic modulation patterns during AVMT and
vibrato conditions. Specifically, respiratory oscillation during AVMT was
hypothesized to correspond with the acoustic extent of SPL modulation whereas
lengthwise oscillation of the vocal folds during vibrato voicing was hypothesized
to correspond with the acoustic extent of fo modulation. The hypothesized
contributions of the respiratory system to SPL modulation were confirmed.
However, the hypothesized contributions of the phonatory system to fo
modulation were not supported. Laryngeal kinematic patterns during vibrato and
AVMT appeared similar although vibrato oscillations occurred at a significantly
higher rate than measured during AVMT. The linkage between the larynx and
respiratory kinematic patterns suggest that the laryngeal movements were
unable to be voluntarily isolated. These findings offer important information
about laryngeal and respiratory physiology that may be further investigated for
utility in the clinical evaluation and treatment of individuals with vocal tremor.
APPENDIX
QUESTIONNAIRE FOR SINGING PARTICIPANTS
60
Questionnaire for Singing Participants
Please answer the questionnaire completely.
Age: _________ Gender: _____________
1) How many years of vocal training have you had?
___________________________________________________________
2) In what genre(s) do you feel most comfortable singing?
___________________________________________________________
3) How would you describe your current singing voice condition (e.g., in-
shape, sing often, been a while since I’ve sung, etc.)?
___________________________________________________________
4) Do you use vibrato while you sing?
___________________________________________________________
a. If you answered yes to #4, how often do you use vibrato while
singing (e.g., all the time, at the end of phrases, etc.)?
_____________________________________________________
b. If you answered yes to #4, have you experienced or are you
currently experiencing any problems with your vibrato? ______
c. If you answered yes to #4, do you vary your vibrato production
across genres? ______
i. If so, how?
________________________________________________
________________________________________________
5) Are you currently experiencing any voice problems? If so, describe:
___________________________________________________________
___________________________________________________________
REFERENCES
Ackermann, H., & Ziegler, W. (1991). Cerebellar voice tremor: an acoustic analysis. Journal of Neurology, Neurosurgery & Psychiatry, 54(1), 74-76. Adler, C. H., Bansberg, S. F., Hentz, J. G., Ramig, L. O., Buder, E. H., Witt, K.,
Edwards, B. W., Krein-Jones, K., Caviness, J. N., (2004). Botulinum toxin type A for treating voice tremor. Archives of Neurology, 61(9), 1416-1420.
Anand, S., Shrivastav, R., Wingate, J. M., & Chheda, N. N. (2012). An acoustic-
perceptual study of vocal tremor. Journal of Voice, 26(6), 811-817. Anand, S., Wingate, J. M., Smith, B., & Shrivastav, R. (2012). Acoustic
parameters critical for an appropriate vibrato. Journal of Voice, 26(6), 819-825.
Barkmeier-Kraemer, J., Lato, K., & Wiley, K. (2011). Development of a speech treatment program for a client with essential vocal tremor. Seminars in Speech and Language, 32(1), 43-57.
Barkmeier-Kraemer, J., & Story, B. (2010). Conceptual and clinical updates on
vocal tremor. ASHA Leader, 15(14), 16-19. Boersma, Paul & Weenink, David (2015). Praat: doing phonetics by computer
[Computer program]. Version 5.4.09, Retrieved 2 June 2015 from http://www.praat.org/
Bové, M., Daamen, N., Rosen, C., Wang, C.-C., Sulica, L., & Gartner-Schmidt, J.
(2006). Development and validation of the vocal tremor scoring system. The Laryngoscope, 116(9), 1662-1667.
Brown, JR & Simonson, J (1963). Organic voice tremor. A tremor of phonation.
Neurology, 13, 520-525. Cohen, S. M., Jacobson, B. H., Garrett, C. G., Noordzij, J. P., Stewart, M. G.,
Attia, A., Cleveland, T. F. (2007). Creation and validation of the singing voice handicap index. Annals of Otology, Rhinology & Laryngology, 116(6), 402-406.
Dalvi, A., & Premkumar, A. (2011). Tremor: etiology, phenomenology, and
clinical features. Disease-a-Month, 57(3), 109-126.
62
Dromey, C., & Smith, M. E. (2008). Vocal tremor and vibrato in the same person: acoustic and electromyographic differences. Journal of Voice, 22(5), 541-545.
Dromey, C., Warrick, P., & Irish, J. (2002). The influence of pitch and loudness
changes on the acoustics of vocal tremor. Journal of Speech, Language & Hearing Research, 45(5), 879.
Farinella K. A., Hixon T. J., Hoit J. D., Story B. H., Jones P. A. (2006). Listener
perception of respiratory-induced voice tremor. American Journal of Speech Language Pathology, 15(1):72-84.
Finnegan, E. M., Luschei, E. S., Barkmeier, J. M., & Hoffman, H. T. (2003).
Synchrony of laryngeal muscle activity in persons with vocal tremor. Archives of Otolaryngology–Head & Neck Surgery, 129(3), 313-318.
Gamboa, J., Jiménez-Jiménez, F. J., Nieto, A., Cobeta, I., Vegas, A., Ortí-Pareja,
M., García-Albea, E. (1998). Acoustic voice analysis in patients with essential tremor. Journal of Voice, 12(4), 444-452.
Gurey, L., Sinclair C., & Blitzer, A. (2013). A new paradigm for the management
of essential vocal tremor with botulinum toxin. Laryngoscope, 123(10), 2497-2501.
Guzman, M. A., Dowdall, J., Rubin, A. D., Maki, A., Levin, S., Mayerhoff, R., &
Jackson-Menaldi, M. C. (2012). Influence of emotional expression, loudness, and gender on the acoustic parameters of vibrato in classical singers. Journal of Voice, 26(5), 675.
Howes, P., Callaghan, J., Davis, P., Kenny, D., & Thorpe, W. (2004). The
relationship between measured vibrato characteristics and perception in Western operatic singing. Journal of Voice, 18(2), 216-230.
Hsiao, T.-Y., Solomon, N. P., Luschei, E. S., & Titze, I. R. (1994). Modulation of
fundamental frequency by laryngeal muscles during vibrato. Journal of Voice, 8(3), 224-229.
Jacobson, B. H., Johnson, A., Grywalski, C., Silbergleit, A., Jacobson, G.,
Benninger, M. S., & Newman, C. W. (1997). The voice handicap index (vhi) development and validation. American Journal of Speech- Language Pathology, 6(3), 66-70.
Jiang, J., Lin, E., & Hanson, D. G. (2000). Acoustic and Airflow Spectral Analysis
of Voice Tremor. Journal of Speech, Language & Hearing Research, 43(1), 191.
63
Kendall, K. A., & Leonard, R. J. (2011). Interarytenoid muscle botox injection for treatment of adductor spasmodic dysphonia with vocal tremor. Journal of Voice, 25(1), 114-119.
Kirkpatrick, A. (2008). Teaching methods for correcting problematic vibratos:
using sustained dynamic exercises to discover and foster healthy vibrato. Journal of Singing, 64(5), 551-556.
Kotby, M., & Fex, B. (1998). the accent method: behavior readjustment voice
therapy. Logopedics Phoniatrics Vocology, 23(1), 39-43. Kotby, M., Shiromoto, O., & Hirano, M. (1993). The accent method of voice
therapy: Effect of accentuations on FO, SPL, and airflow. Journal of Voice, 7(4), 319-325.
Lederle, A., Barkmeier-Kraemer, J., & Finnegan, E. (2012). Perception of vocal
tremor during sustained phonation compared with sentence context. Journal of Voice, 26(5), 668-669.
Lester, R. A., Barkmeier-Kraemer, J., & Story, B. H. (2013). Physiologic and
acoustic patterns of essential vocal tremor. Journal of Voice, 27(4), 422-432.
Louis, E. D., & Machado, D. G. (2015). Tremor-related quality of life: A
comparison of essential tremor vs. Parkinson's disease patients. Parkinsonism & Related Disorders, 21(7), 729-735.
Lowell S. Y., Barkmeier-Kraemer J. M., Hoit J. D., Story B. H. (2008).
Respiratory and laryngeal function during spontaneous speaking in teachers with voice disorders. Journal of Speech, Language, and Hearing Research, 51(2):333-49.
Lundy, D. S., Roy, S., Xue, J. W., Casiano, R. R., & Jassir, D. (2004).
Spastic/spasmodic vs. tremulous vocal quality: motor speech profile analysis. Journal of Voice, 18(1), 146-152.
Pettersen, V., & Westgaard, R. H. (2005). The activity patterns of neck muscles
in professional classical singing. Journal of Voice, 19(2), 238-251. Prame, E. (1994). Measurements of the vibrato rate of ten singers. The Journal
of the Acoustical Society of America, 96(4), 1979-1984. Ramig, L. A., & Shipp, T. (1987). Comparative measures of vocal tremor and
vocal vibrato. Journal of Voice, 1(2), 162-167. Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland,
USA, Retrieved from http://imagej.nih.gov/ij/, 2015
64
Schneider, S. A., & Deuschl, G. (2015). Medical and surgical treatment of tremors. Neurologic Clinics, 33, 57-75.
Seashore, C. E. (1931). The natural history of the vibrato. Proceedings of the
National Academy of Sciences of the United States of America, 17(12), 623-626.
Story, B. H. (1995). Physiologically-based speech simulation using and
enhanced wave-reflection model of the vocal tract. (Unpublished doctoral dissertation). University of Iowa, Iowa City, IA.
Sulica, L., & Louis, E. D. (2010). Clinical characteristics of essential voice tremor:
A study of 34 cases. The Laryngoscope, 120(3), 516-528. Sundberg, J. (1994). Acoustic and psychoacoustic aspects of vocal vibrato. STL-
QPSR, 35(2-3), 045-068. Taha, J. M., Janszen, M. A., & Favre, J. (1999). Thalamic deep brain stimulation
for the treatment of head, voice, and bilateral limb tremor. Journal of Neurosurgery, 91(1), 68-72.
Titze, I. R., Story, B., Smith, M., & Long, R. (2002). A reflex resonance model of
vocal vibrato. The Journal of the Acoustical Society of America, 111(5), 2272-2282.
Tomoda, H., Shibasaki, H., Kuroda, Y., & Shin, T. (1987). Voice tremor:
dysregulation of voluntary expiratory muscles. Neurology, 37(1), 117-122. Warrick, P., Dromey, C., Irish, J., Durkin, L., Pakiam, A., Lang, A. (2000).
Botulinum toxin for essential tremor of the voice with multiple anatomical sites of tremor: a crossover design study of unilateral versus bilateral injection. Laryngoscope, 110(8), 1366-74.
Watson, A. H. D., Williams, C., & James, B. V. (2012). Activity patterns in
latissimus dorsi and sternocleidomastoid in classical singers. Journal of Voice, 26(3), 95-105.
Wolraich, D., Marchis-Crisan, C.V., Redding, N., Khella, S.L., & Mirza, N. (2010).
Laryngeal tremor: co-occurrence with other movement disorders. Otorhinolaryngology, 72, 291–294.
